Stats on compact block reconstructions

Prefilling, a feature added for convenience in the design of compact blocks, has shown to be more problematic than beneficial. It was intended to streamline transactions but instead introduces several issues. One major problem is that it can delay the reception of compact blocks when not necessary. Since peers receive compact blocks from multiple sources, the use of prefill by all these sources can lead to a significant waste of bandwidth. This wastage is compounded by the fact that transactions missed are often too large, exacerbating the inefficiencies caused by prefilling.

Fiber, despite being licensed under AGPL, offers solutions to these challenges. It allows for the transmission of additional data which can improve efficiency even when predictions on missed transactions are inaccurate. Fiber facilitates data contribution from multiple peers and proposes the use of UDP to overcome TCP window limitations. However, the deployment of UDP poses its own set of challenges, notably the requirement for hole punching, which may hinder widespread adoption.

Recent developments have seen the integration of minisketch into the core, although it remains unused. Utilizing such tools has demonstrated potential for significantly reducing compact block size, thereby providing ample space for missing transaction data. Nevertheless, the effectiveness of these solutions may be limited to scenarios involving peers with high bandwidth capabilities that can manage congestion manually. The inclusion of large amounts of data that were never relayed by miners raises doubts about the overall performance of these schemes. Nonetheless, the fact that improvements can be realized even with partial adoption among nodes highlights the potential for innovative development and deployment strategies in network protocols.